Condition monitoring system for tools in machining using machine learning algorithm

Automatic machining is an integral and important part of modern manufacturing systems. One of the most significant issues in controlling and optimizing the automatic machining process is the detection of tool wear and failure during the process. Awareness of tool status during the process is only possible by establishing a precise and efficient condition monitoring system. The objective of this research is to investigate the feasibility of creating a simple, accurate, and efficient tool condition monitoring system using a machine learning algorithm for single-point cutting tools with a flank wear criterion. To achieve this goal, a novel electromechanical impedance method and machine learning decision tree algorithm were employed. In the electromechanical impedance method, the mechanical impedance of the structure is measured using an electrical equivalent, and tool monitoring is performed using data from a piezoelectric sensor (simultaneous actuator/sensor) installed on the tool, analyzing changes in the tool’s impedance. In the machine learning method, the computer receives the dataset, designs the algorithm, and trains it. In this study, a supervised machine learning method using a decision tree was utilized. The results indicate that (in cases where experimental data is limited), using the k-fold method in the tool condition monitoring system performs better than the hold-out method. However, careful selection of the value of k is crucial, as choosing an appropriate k significantly impacts the performance of the machine.

Condition monitoring system for tools in machining using machine learning algorithm Keywords:

Machining , machine learning , decision tree , tool condition